What this calculator does
A receiver or transmitter line-up is a chain of two-ports: cables, filters, amplifiers, mixers and attenuators. Each stage has a gain, a noise figure, a third-order intercept and a compression point. The cascade analysis turns those per-stage numbers into the system figures you actually specify: total gain, noise figure, input and output IP3, P1dB, noise floor, spurious-free dynamic range (SFDR) and sensitivity. It also shows the signal and noise level at every stage output, so you can see where the line-up runs out of headroom.
The table above takes 1 to 20 stages. All the arithmetic is done in linear units (mW and power ratios), and the results are converted back to dB and dBm only for display. Stages are evaluated in the order listed, from the antenna (or signal source) towards the output.
Gain and noise figure (Friis)
The cascade gain is the sum of the stage gains in dB. The noise factor follows Friis's formula, in which each stage's excess noise is divided by the gain ahead of it:
Here is the reference temperature that defines noise figure. The NF share column shows each term as a percentage of the total . This percentage is also the stage's share of the cascade noise temperature, so it tells you which stage to improve first.
Passive stages. A matched lossy stage (cable, filter, attenuator) at physical temperature has a noise figure equal to its loss, so . When you choose Filter / attenuator or Cable, the NF column is locked to the loss and you enter the loss as a negative gain. One consequence is that a loss of dB in front of the cascade adds exactly dB to the system NF. The test suite checks this property.
Intercept points: coherent vs. RSS
Each stage's IP3 can be entered as an input (IIP3) or output (OIP3) value; the two are related by . Referred to the cascade input, the intercepts combine as follows.
Coherent (default, worst case). The third-order products of all stages are assumed to add in phase as voltages:
This is the formula in Pozar, Egan and Razavi, and it is the basis most specifications are written against. It never overestimates the intercept of a cascade of memoryless stages.
RSS (incoherent). If the IM3 products of different stages have unrelated phases, their powers add instead of their voltages:
RSS always gives a higher IIP3 than coherent summation. When two stages contribute equally, the difference is dB. When one stage dominates, the two methods converge. Use RSS as an optimistic bound or a "typical" estimate. Filters with phase shift between stages, or wideband signals, usually measure somewhere between the two. The IP3 share column gives each stage's term as a percentage of the sum (or of the sum of squares in RSS mode). A stage with a high share is the one limiting linearity.
Compression (P1dB)
The cascaded P1dB uses the same summation on input-referred compression points. A stage's input P1dB is , because at the 1 dB point the gain has dropped by 1 dB. The cascade output value is . Some spreadsheets omit the 1 dB, which makes their input P1dB 1 dB lower than the value shown here.
This is an approximation. Compression is a large-signal effect, and the real cascade 1 dB point depends on the full AM/AM curve of every stage. In practice the last active stage usually dominates, and the estimate is good to about a dB when one stage clearly sets the limit. Treat it as a planning number and confirm it by measurement or harmonic-balance simulation.
The level diagram and the Headroom column compare each stage's small-signal output level with its own OP1dB. Stages within the compression warning margin (6 dB by default) are flagged.
Noise floor, SFDR and sensitivity
The noise floor is referred to the cascade input:
With the source at , this reduces to the familiar (more precisely dBm/Hz). For an antenna looking at cold sky, set to the antenna temperature. The noise figure itself is always referred to 290 K.
From the floor and the input intercept, the calculator gives:
- SFDR . This is the range of input levels over which two equal tones produce third-order products no larger than the noise floor.
- Sensitivity (MDS) . This is the weakest signal that reaches the required SNR in the noise bandwidth .
Use the noise bandwidth of the final channel filter, not its 3 dB bandwidth, if you know it.
Worked example
The default line-up is a superheterodyne front end with a 1 MHz noise bandwidth, an input signal of −50 dBm and a source at 290 K:
| # | Stage | Gain | NF | IP3 | OP1dB |
|---|---|---|---|---|---|
| 1 | Antenna cable | −0.5 dB | 0.5 dB (= loss) | ideal | ideal |
| 2 | LNA | 18 dB | 1.2 dB | OIP3 +32 dBm | +20 dBm |
| 3 | Band-pass filter | −2 dB | 2 dB (= loss) | ideal | ideal |
| 4 | Mixer | −7 dB | 7 dB | OIP3 +15 dBm | +3 dBm |
| 5 | IF amplifier | 20 dB | 3 dB | OIP3 +35 dBm | +20 dBm |
Noise. The gains ahead of the five stages are 1, 0.891, 56.23, 35.48 and 7.079. The Friis terms are 0.1220, 0.3571, 0.0104, 0.1131 and 0.1406, which sum to . So , giving NF = 2.41 dB and K. The LNA contributes 48.0 % of the excess noise. The 0.5 dB cable in front of it contributes 16.4 %, because nothing amplifies the signal before the cable.
Linearity. The stage input intercepts are +14 dBm (LNA), +22 dBm (mixer) and +15 dBm (IF amplifier). Referred to the antenna, the coherent terms are 0.0355, 0.2239 and 0.2239 mW⁻¹. Their sum is 0.4832 mW⁻¹, so IIP3 = +3.16 dBm and OIP3 = +31.66 dBm. The mixer and the IF amplifier each account for 46.3 %: with this much gain in front of them, they limit linearity, not the LNA. With the RSS toggle, IIP3 rises to +4.97 dBm.
Compression. The estimated input P1dB is −9.49 dBm and the output P1dB is +18.01 dBm.
Floor and dynamic range. −111.56 dBm. With a 10 dB SNR requirement, the sensitivity is −101.56 dBm. SFDR 76.48 dB. The −50 dBm input leaves the cascade at −21.5 dBm, more than 40 dB below the IF amplifier's +20 dBm OP1dB.
Accuracy and validity
| Quantity | Model | Error sources |
|---|---|---|
| Gain, NF | Exact for matched stages (available gain) | Mismatch between stages; NF quoted at a different source impedance |
| IIP3 / OIP3 | Coherent: exact worst case for memoryless 3rd-order stages | Frequency-dependent phase (filters), higher-order terms |
| P1dB | Estimate from intercept-style summation | Typically ±1 dB when one stage dominates |
| Noise floor, MDS | Noise bandwidth vs. 3 dB bandwidth; image noise in mixers |
The tests check the results against an independent Python implementation of the textbook formulas, written separately from the TypeScript (it works with noise temperatures and output-referred intercepts), on 24 random cascades of 1 to 10 stages. They also check exact limits: a single stage is unchanged, an ideal 0 dB stage changes nothing, and a pad in front adds its loss to NF and IIP3.
Practical notes
- Use available gain. Datasheet differs from available gain when stages are mismatched. For a well-matched 50 Ω chain the difference is small; across a filter skirt it is not.
- Mixer NF. Use the SSB noise figure for a single-sideband receiver. A DSB figure is about 3 dB optimistic for an image-rejected design.
- Put the loss after the gain. Every dB of loss in front of the LNA appears directly in the NF. Behind 20 dB of gain, the same loss costs a few hundredths of a dB.
- Gain is not free. Extra gain ahead of the mixer lowers NF but raises the IM3 share of every later stage. The contribution chart shows that trade-off.
- Cross-check. The CSV export gives every per-stage value to 0.001 dB for comparison with a spreadsheet.
Sources for the formulas
Friis (1944) for the noise cascade; Pozar, Microwave Engineering (4th ed., ch. 10) for passive-stage noise, intercept points and dynamic range; Egan, Practical RF System Design, for coherent and RSS intermodulation and compression in line-ups; Razavi, RF Microelectronics (2nd ed., ch. 2) for cascaded nonlinearity. The full citations are listed below.